Resumen
The optimization of the breast cancer diagnostics in considered in this paper. The issue of classifying a tumor as benign or malignant is being resolved. Histochemistry of breast tumor punctates is performed using a fine-needle aspiration puncture. Next, a digital analysis of images of cell nuclei in micrographs obtained from sections of tissue pieces is performed. Working within the framework of the standard approaches of medical diagnostics, the authors carried out two stages: diagnostic features ranking was done at the first stage, and then patients classification was fulfilled at the second stage. A discrete error function was used to register the number of incorrectly diagnosed patients. The authors studied the geometric characteristics of the behavior of a discrete error function in the parameter space of a linear discriminant function. We developed and programmed an algorithm for movement to a minimum, visually showing in 2D and 3D the dynamics of changes in the error function values. As a result, it was shown that a sufficiently reliable diagnosis with an accuracy of 94.73% can be carried out based only by two diagnostic features.